CITI Talk: “Maximising the Utility of Virtually Sliced Millimetre-Wave Backhauls via a Deep Learning Approach”, Rui Li, PhD student at the University of Edinburgh, Inria antenne

Title

Maximising the Utility of Virtually Sliced Millimetre-Wave Backhauls via a Deep Learning Approach

Abstract

Advances in network programmability enable operators to ‘slice’ the physical infrastructure into independent logical networks. By this approach, each network slice aims to accommodate the demands of increasingly diverse services. Precise allocation of resources to slices across future 5G millimetre-wave backhaul networks, so as to optimise their utility, is however challenging. This is because the performance of different services often depends on conflicting requirements, including bandwidth, sensitivity to delay, or the monetary value of the traffic incurred. In this talk, I will present our recent work in which we propose a general rate utility framework for slicing mm-wave backhaul links, which encompasses all known types of service utilities, i.e. logarithmic, sigmoid, polynomial, and linear. We then employ a deep learning solution to tackle the complexity of optimising non-convex objective functions built upon arbitrary combinations of such utilities. Specifically, using a stack of convolutional blocks, our approach can learn correlations between traffic demands and achievable optimal rate assignments. The proposed solution can be trained within minutes, following which it computes rate allocations that match those obtained with state-of-the-art global optimisation algorithms, yet orders of magnitude faster. This confirms applicability to highly dynamic traffic regimes and we demonstrate up to 62% network utility gains over a baseline greedy approach.


PhD position on Mobile Crowd Sensed Data Analysis (Agora team – CITI Lab – INRIA/INSA-Lyon)

Topic

Mobile Crowd Sensed Data Analysis: Application to Participatory Environmental Crowd Sensing in Smart Cities

Thesis Description

The growing emergence of low-cost environmental monitoring systems combined with the recent advances in the design of Internet-of-Things architectures and protocols has given a new impetus to smart cities applications which is expected to significantly enhance the fine-grained characterization of different physical quantities in our cities (air quality, temperature, noise, etc.).

In this perspective, a promising approach is to involve citizens in the monitoring process using low-cost platforms and built-in sensors in order to collectively monitor different physical quantities. While relying on very high number of people to gather data is promising in accumulating large volumes of data, issues such as dealing with the variation in data accuracy due to the heterogeneity of sensing hardware and conditions, space-time continuity of measures, phenomena dynamics, impact of mobility on sensor quality, etc. arise and make it challenging to efficiently analyse the mobile crowd sensed data [1].

The aim of this thesis is to propose and evaluate novel solutions for efficient fine-grained mapping of physical phenomena based on mobile crowd sensed data with a focus on air quality and temperature.  Two directions will be explored. The first direction is based on data interpolation using techniques such as log-normal regression [2], deep learning [2,3], generalized additive modelling [4,5], Kriging-based modelling [6], etc. The second direction concerns data assimilation where the measures are incorporated into numerical models of the studied phenomena [7,8].

The Ph.D. student is expected to design novel solutions and conduct mathematical analysis on them. The validation and evaluation of the proposed solutions should include comparisons with state-of-the-art proposals. Data used in these evaluations is expected to come from the results of several participatory planned measurement campaigns.

Thesis context

This thesis is part of the 3M’Air multidisciplinary project, funded by the cluster of excellence IMU (LabEx Intelligence of Urban Worlds).  The 3M’Air project aims to study the potential of participatory crowd sensing to improve fine-grained knowledge of air quality and temperature while dealing with main scientific, technological, geographical and sociological issues. For that purpose, 3M’Air brings together the scientific and technical skills of three research laboratories: CITI (on wireless communications and data analysis), LMFA (on fluid mechanics and urban atmospheric dispersion models) EVS (on geographical and sociological issues) as well as five other operational partners: ATMO-Aura (the regional air quality observatory), Meteo France  (The French national meteorological service), le Grand Lyon (Greater Lyon urban community), Ville de Lyon (City of Lyon) and Lyon Meteo (a local company working on meteorological services).

The successful candidate will join the INRIA research group Agora located in Lyon, which is part of the CITI laboratory. The thesis will be mainly co-supervised by Dr. Walid Bechkit and Prof. Hervé Rivano of the CITI lab with a strong collaboration with the LMFA and the EVS laboratories.

Applications

Interested candidates should send a detailed CV with information on education, obtained degrees and qualification, as well as a cover letter detailing the motivation and scientific background of the candidate. Applications should also include the names and contact details of two referees.

Applications should be submitted by email to: walid.bechkit@inria.fr cc hervé.rivano@inria.fr, a rolling deadline applies.

Some references

[1] M. Fiore, A. Nordio and C-F. Chiasserini, “Driving Factors Toward Accurate Mobile Opportunistic Sensing in Urban Environments”, IEEE Transactions on Mobile Computing, Vol. 15, pp. 2480–2493, 2016

[2] A. Marjovi, A. Arfire, and A. Martinoli, “Extending Urban Air Quality Maps Beyond the Coverage of a Mobile Sensor Network: Data Sources, Methods, and Performance Evaluation”, in proc. of EWSN, pp. 12-23, 2017.

[3] M. D. Adams and P. S. Kanaroglou “Mapping real-time air pollution health risk for environmental management: Combining mobile and stationary air pollution monitoring with neural network models”, Journal of environmental management, vol. 168, pp. 133-141, 2016.

[4] D. Hasenfratz, O. Saukh, C. Walser, C. Hueglin, M. Fierz, T. Arn, J. Beutela and L. Thielea, “Deriving high-resolution urban air pollution maps using mobile sensor nodes”, Pervasive and Mobile Computing, vol. 16, pp. 268-285, 2015.

[5] M. Mueller, D. Hasenfratz, O. Saukh, M. Fierz, and C. Hueglin, “Statistical modelling of particle number concentration in Zurich at high spatio-temporal resolution utilizing data from a mobile sensor network”, Atmospheric Environment, vol. 126, pp. 171-181, 2016.

[6] V. Singh, C. Carnevale, G. Finzi, E. Pisoni, and M. Volta, “A cokriging based approach to reconstruct air pollution maps, processing measurement station concentrations and deterministic model simulations”, Environmental Modelling & Software, vol. 26, pp. 778-786, 2011.

[7] A. Tilloy, V. Mallet, D. Poulet, C. Pesin, and F. Brocheton, “BLUE-based NO2 data assimilation at urban scale”, Journal of Geophysical Research: Atmospheres,  vol. 118, pp 2031-2040, 2013.

[8] A. Boubrima, W. Bechkit, H. Rivano and L. Soulhac. “Leveraging the Potential of WSN for an Efficient Correction of Air Pollution Fine-Grained Simulations”, to appear in proc. of IEEE ICCCN 2018.


CITI Talk: “Recycler les ondes radio ambiantes pour connecter les objets”, Dinh-Thuy PHAN-HUY (Orange, Chatillon), 22 May (10h30 in TD-C)

Titre

Recycler les ondes radio ambiantes pour connecter les objets

Description

o   Lors de l’édition 2017 du Salon de la recherche Orange, du 5 au 7 décembre (https://hellofuture.orange.com/fr/lenergy-free-communication-donne-des-ailes-aux-objets-connectes/), Orange a réalisé, pour la première fois, une transmission de données sans fil,  effectuée grâce aux seules ondes déjà diffusées par… la tour Eiffel ! Aucune onde supplémentaire n’a été émise. Cette technologie dite de rétro-diffusion ambiente découverte par l’Université de Washington en 2013, a une sobriété énergétique exceptionnelle. Elle permet de fournir de nouveaux services sans dépenser plus en spectre et en puissance rayonnée, ouvre d’énormes possibilités en termes d’utilisation massive d’objets connectés pour les villes, les maisons et les usines intelligentes.

o   Aujourd’hui, pour la  première fois, le projet ANR SpatialModulation (https://spatmodulation.cms.orange-labs.fr/) dirigé par Orange, tentera une démonstration en temps réel d’une communication utilisant les ondes TV de la Fourvière, entre un « émetteur » (qui n’émet pas) développé par Orange et un récepteur développé par l’Institut Langevin sur GNU Radio.


CITI Talk: “Optimization Algorithms for Solving Problems Arising from Large Scale Machine Learning”, Vyacheslav Kungurtsev (Czech Technical University, Prague), May 7th, at 2pm in “salle TD-C” ( Claude Chappe Building)

Title

Optimization Algorithms for Solving Problems Arising from Large Scale Machine Learning

Abstract

In the contemporary “big data” age, the use of Machine Learning models for analyzing large volumes of data has been instrumental in a lot of current technological development. These models necessitate solving very large scale optimization problems, presenting challenges in terms of developing appropriate solvers. In addition, especially for problems arising from Deep Neural Network architectures, the resulting problems are often nonconvex, and sometimes nonsmooth, giving additional difficulty.

In this talk I present the standard structural elements of this class of problems, and how these structures can be handled with appropriate parallel architectures. I discuss the state of the art in terms of optimization algorithms for this setting and summarize the prognosis for ongoing and future research.